[FFmpeg-devel] [PATCH] Re-add OS/2 support
Tue Sep 11 07:03:40 CEST 2007
Diego Biurrun wrote:
> On Sun, Sep 09, 2007 at 10:48:38AM -0700, Dave Yeo wrote:
>> Diego Biurrun wrote:
>>> Patch in its current form is rejected.
>> Lets try a different approach. Attached is the patch that I've been using.
>> As long as the SDL dlls are fairly new we no longer need the MorphToPM
>> stuff. Also need fairly new SDL and fslib for fullscreen to work.
> Then we should just require new SDL stuff IMO.
Yes, I agree. Also KO Myung-Hun has an nice adaption of ffplay which
instead of SDL uses native DART and DIVE.
Actually I am now having some minor problems with ffplay. Doesn't exit
until I press q and full screen is screwed up. Quite likely my SDL
install is broken, couple of sets of headers and import libs :)
>> make test creates files in $TMP then can fail as it egreps /tmp
> That is a separate issue, patches welcome.
>> --- ffmpeg/configure Sun Sep 9 00:01:48 2007
>> +++ ffmpeg.os2/configure Sun Sep 9 08:55:20 2007
>> @@ -1196,6 +1199,26 @@
>> ranlib="echo ignoring ranlib"
>> + os/2)
>> + FFSERVERLDFLAGS=""
> ffserver works?
It seems to, haven't particularly tested it much. The server regression
test fails but it also says that ffserver is broken and the regression
tests fail randomly. For OS/2 it is broken in the sense that it looks
for its config files in /etc instead of $ETC, separate issue of course.
>> + LIBPREF=""
>> + LIBSUF="_s.lib"
>> + SLIBPREF=""
>> + SLIBSUF=".dll"
>> + SLIBNAME_WITH_MAJOR='$(SLIBPREF)$(NAME)-$(LIBMAJOR)$(SLIBSUF)'
This just fixes the DLL name so instead of foo.dll.xx we get foo.xx.dll
where xx is the major number. Thanks to the mingw porter for this.
>> + SLIB_EXTRA_CMD="dllar -o \$(NAME)\$(SLIBSUF) \$(NAME)\$(LIBSUF) -d \"\$(SLIBNAME_WITH_MAJOR)\" -omf -ord \$(LDFLAGS) \$(EXTRALIBS) \$(EXTRAOBJS); cp -f \$(SLIBNAME) \$(SLIBNAME_WITH_MAJOR)"
This passes the static lib to dllar.cmd
a REXX script distributed with GCC which basically extracts the imports
from the static lib, creates a DEF file, links the object files and DEF
files into a DLL and also creates the import libs in both a.out and OMF
format. The -d is for the description in the DEF file and DLL and the
-omf gives OMF support and -ord gives ordinals to help enforce
versioning and a small speedup. We then copy the short named DLL over
the long named DLL.
This is a bit wasteful as the build process builds a broken DLL then
overwrites it but seemed simplest without making large changes in
configure and the makefiles.
>> + SLIB_INSTALL_EXTRA_CMD="-install -m 644 \$(SLIBNAME:\$(SLIBSUF)=.lib) \$(SLIBNAME:\$(SLIBSUF)=.a)"
This just installs the import libs, both OMF and a.out format.
> Maybe you can explain the tricks you are playing here...
> Apart from that the patch looks OK to me. Mans?
I also just realized that make install was slightly broken when built
with --enable-shared. Updated patch included. (SLIB_INSTALL_EXTRA_CMD
was missing target and added SLIBNAME_WITH_VERSION).
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 3259 bytes
Desc: not available
More information about the ffmpeg-devel